Multiomics Study of Key Differentially Expressed Methylation Genes in HCC Based on Epigenome and Transcriptome Analyses

Objective: To screen out key differentially expressed methylation genes in hepatocellular carcinoma (HCC) by extracting data from the NCBI-GEO and TCGA databases, discuss the interaction of key genes in HCC and explore the inuence of these genes on tumor progression. Methods: The clinical information and sequence data of HCC patients and healthy controls were extracted from the NCBI-GEO and TCGA databases. The module that presented the highest correlation with the tumor phenotypes was selected by the WGCNA network and TOM analysis. GO and KEGG analyses were used for signaling pathway analysis. Key differentially expressed genes were screened out from the module. KM plot analysis was performed to determine the impact of key genes on patient survival. The relationship between core genes and tumor immune inltration was also discussed. Drug sensitivity analysis was performed to observe the sensitivity of key genes methylated by chemotherapeutic drugs and thus determine the associated effects on patient prognosis. Finally, the molecular mechanisms of the genes involved in tumor progression were analyzed by GSVA, and the interactions between key genes were revealed using Gene MANIA analysis. Results: A total of 373 differentially expressed genes were screened, including 88 upregulated genes and 285 downregulated genes. Three of these genes, CHST4, CRHBP, and IGFBP3, were found to be abnormally methylated in HCC patients and signicantly associated with expected survival. The relationship between these key genes and tumor immune invasion was conrmed, and these genes play a protumor or antitumor role through immune cell mediation. In addition, we found that CHST4 and IGFBP3 expression can signicantly reduce the sensitivity of several commonly used chemotherapy drugs to tumor cells, while CRHBP expression plays a synergistic role with chemotherapy drugs to enhance mutual effects. The participation of these genes in specic signaling pathways involved in liver metabolism and tumor progression has also been identied. Finally, complex interactions between the three core genes and other related genes were presented. Conclusion: The CHST4, CRHBP and IGFBP3 genes present signicant methylation and differential expression in HCC, are involved in tumor immune inltration, and affect chemotherapy drug sensitivity. As potential biomarkers and therapeutic targets, they may be benecial to the diagnosis and prognosis of HCC patients in the future.

Introduction HCC is an extremely dangerous and common malignant tumor originating from hepatocytes and accounts for approximately 80% of all primary liver cancers. In 2018, HCC presented the sixth highest incidence among cancers and was the fourth leading cause of cancer death in the world, thus contributing to 841,000 new cases and 782,000 deaths annually [1]. HCC is generally considered to present a lengthy development process caused by various complex factors, including chronic hepatitis B virus (HBV) and hepatitis C virus (HCV) infections, a atoxin B1, alcoholism, smoking, water pollution, metabolic syndrome, heredity factors and medication [2,3]. Signi cant regional differences are observed in the distribution of HCC. The majority of adult cases (80%) occur in South Africa and East Asia, where people have long suffered from hepatitis and a atoxin B1. However, HCV and excessive alcohol consumption are major risk factors in the USA and Europe. In addition, surgical di culty and poor prognosis are commonly observed for HCC due to its late diagnosis and high recurrence rate [2,4,5]. Currently, the combination of liver ultrasound and alpha-fetoprotein (AFP) is still the main method of identifying HCC at the very early stage. Unfortunately, the reported sensitivities were only 63% when imaging and serology were adopted simultaneously. Even HCC nodules detected early carry up to a 70% risk of ve-year recurrence after radical resection [6]. Therefore, more accurate detection and more effective treatment protocols are urgently needed.
The occurrence of HCC is considered a negative outcome of chronic interactions involving multiple pathways. Although the speci c pathogenesis of HCC remains a mystery, genetic and epigenetic abnormalities have been con rmed to exist in most tumor cells, and they include genome deletion, ampli cation, insertion, and abnormal DNA methylation [7][8][9]. DNA methylation in eukaryotes refers to the formation of 5-methylcytosine via the addition of a methyl group to the carbon-5 position of cytosine in the CpG dinucleotide sequence. This modi cation can inhibit downstream gene expression without changing the nucleotide order and structure. DNA methylation is an important factor in cell differentiation and proliferation. Abnormal methylation of tumor suppressor genes and demethylation of oncogenes are key events in tumorigenesis and progression, including HCC. Abnormal DNA methylation is triggered when risk factors, such as alcohol, HBV or HCV, impact the liver to initiate the in ammatory response.
Once abnormalities accumulate to a certain extent, the normal cells of the liver will become tumorigenic, thereby accelerating the development of HCC [10]. DNA methylation abnormalities usually occur at the embryonic stage of HCC, even before the appearance of HCC [11,12], and they are potentially reversible [10]. Therefore, it is of great clinical signi cance to identify differentially expressed methylation genes and investigate the regulatory mechanism of tumor cells at the epigenetic level because such information could help identify new biomarkers and effective targeted therapies, which may lead to early screening and improved long-term survival.
In this article, we integrated the data collected from the National Center for Biotechnology Information Gene Expression Omnibus (NCBI-GEO) and The Cancer Genome Atlas (TCGA). The key differentially expressed methylation genes were screened by a standard methylation process analysis and weighted gene coexpression network analysis (WGCNA). Then, veri cation experiments were carried out to con rm the conclusions. In summary, our study will provide insights for future breakthroughs in targeted therapies for HCC.
Material And Methods 1. Data download NCBI-GEO (https://www.ncbi.nlm.nih.gov/geo/) is a gene expression database that stores chip and highthroughput sequencing data created and maintained by the NCBI. We downloaded the series matrix le of the GSE57957 dataset from the NCBI-GEO public database, and it contained transcriptome data from 78 groups, including healthy controls (n=39) and HCC patients (n=39). A difference analysis was performed, and the selection criteria were |LogFC|>1 and P-value <0.05. Similarly, methylation data for the series matrix le of the GSE113017 dataset, including healthy controls (n=30) and HCC patients (n=30), were downloaded from NCBI-GEO. The CHAMP function package was used for a 450K differential methylation site analysis. The screening required that one group had a β-value less than 0.2 and the other group had a β-value greater than 0.3. An adjusted P-value <0.05 was adopted as the cutoff criterion.

GO and KEGG pathway analysis
Gene Ontology (GO) is known for its ability to de ne and describe the function of genes in molecular function, biological process and cellular components categories. Kyoto Encyclopedia of Genes and Genomes (KEGG) is a comprehensive database that integrates genomic, chemical and systematic information, with an emphasis on storing information on gene pathways in different species. The Metascape database (www.metascape.org) was utilized to annotate and visualize the biological functions and signaling pathways involved in the development of HCC. Then, GO and KEGG pathway enrichment analyses were performed for the selected speci c genes. Min overlap values ≥3 and P-values ≤0.01 were considered statistically signi cant.

WGCNA
WGCNA is one type of analysis that is performed to identify coexpression gene modules and explore the location of core genes and the association between genes and phenotypes by constructing a weighted gene coexpression network. The WGCNA-R package was used to construct the coexpression network for all the genes contained in the dataset. Then, we screened the top 5,000 genes presenting variance with this algorithm for further analysis. The soft threshold was set as 6. For the estimation of network connectivity, a weighted adjacency matrix was transformed into a topological overlap matrix (TOM). Moreover, a hierarchical clustering method was utilized to construct the clustering tree structure of the TOM. The different modules were colored differently to distinguish the different gene modules on the branches of the cluster tree. Based on the weighted correlation coe cient, the screened genes were classi ed according to the expression pattern. Genes with similar patterns can be grouped into a single module; therefore, 5000 genes were divided into multiple regular modules through gene expression patterns.
4. Data collection from TCGA TCGA (https://portal.gdc.cancer.gov/) is the largest known cancer gene information database and stores large amounts of data, including data on gene expression, miRNA expression, copy number variation, DNA methylation, SNP, etc. We downloaded the processed original data of mRNA expression in HCC from the TCGA database. A total of 369 patient specimens with complete expression pro les and clinical information were collected for follow-up analysis.

Drug sensitivity analysis
Genomics of Drug Sensitivity in Cancer (GDSC, https://www.cancerrxgene.org/) is the largest database of pharmacogenomics, and it was used in conjunction with the R software package "pRRophetic" to predict the chemotherapy sensitivity of each tumor sample. The estimated IC 50 for each speci c chemotherapeutic agent was obtained by regression analysis. Regression and prediction accuracy were guaranteed by 10 cross-validation tests with the GDSC training set. Default values were selected for all parameters, including ComBat with removing batch effects and average repeated gene expression. .

GSVA
Gene set variation analysis (GSVA) is a nonparametric and unsupervised method of evaluating the enrichment of transcriptome genomes, and it converts changes from the gene level to the pathway level to facilitate the determination of the biological function in the sample by comprehensively scoring gene sets of interest. In this study, gene datasets were downloaded from the Molecular Signatures Database (Version 7.0). GSVA was adopted to comprehensively score each gene dataset for the evaluation of potential biological function changes among different samples.

GeneMANIA analysis
GeneMANIA (http://www.genemania.org) is a exible and user-friendly PPI network-building database for visualizing gene functional networks and analyzing functions and interactions among related genes. The website can set up the data source of gene nodes. Additionally, it includes a variety of bioinformatics analysis methods, such as physical interaction, gene coexpression, gene colocation, gene enrichment analysis and website prediction. In this search, the interaction network of core genes formed by GeneMANIA was used to explore and explain the possible mechanism impacting the core genes in HCC patients.
. Statistical analysis All statistical analyses were conducted using R (Version 3.6). All statistical tests were bilateral. A P-value <0.05 was considered statistically signi cant.

Module selection
First, we downloaded transcriptome data for 78 patient groups in the GSE57957 dataset from the NCBI-GEO public database, including healthy controls (n=39) and HCC patients (n=39). The WGCNA network was constructed to explore and screen the key regulatory genes in HCC according to the clinical characteristics of patients. For comparison, HCC patient groups were marked as orange while control groups were marked as white (Fig. 1A). The soft threshold β-value was determined by sft$PowerEstimate.
We selected the most accurate value greater than the baseline of 0.9 for scale independence and close to 0 for mean connectivity. Therefore, the soft threshold was set as 6 ( Fig. 1B). Then, 10 gene modules were detected based on the TOM analysis, including black (225), blue (1261), brown (763), green (241), gray (199), magenta (98), pink (200), red (234), turquoise (1316) and yellow (463) modules. We further analyzed the relationship between modules and traits and found that the magenta module had the highest module-trait relationship (cor=-0.74, p=7e −15 ) (Fig. 2C). Moreover, the relationship between module membership and gene signi cance was signi cant in the magenta module (cor=0.78, p=3e -21 ) (Fig. 2D). An additional le shows this in more detail (see Additional le 1). Therefore, the magenta module was chosen for subsequent validation.

Signaling pathways by enrichment analysis
Once the magenta module was selected, all genes in the module were analyzed by GO and KEGG. Our results showed that the genes of the magenta module were mainly enriched in the following mechanisms closely related to the occurrence and progression of HCC: collagen-containing extracellular matrix, complement activation, lectin pathway, blood vessel development, cytokine receptor binding and other signaling mechanisms (Fig. 2.1.A-B). This outcome was consistent with our expectation, suggesting that the mechanism of the magenta module that affects HCC progression is intimately related to the pathways mentioned above. The interaction between genes in the module is shown in the PPI diagram ( Fig. 2.2).

Differentially methylated genes in HCC
We downloaded the GSE113017 methylation dataset (450K chip) with 60 groups of volunteers from the NCBI-GEO public database, including healthy control groups (n=30) and HCC patient groups (n=30).
Subsequently, the standard methylation process analysis was performed using the CHAMP package. The results showed that a total of 4614 differentially methylated sites were screened, and they included 3838 upregulated sites and 776 downregulated sites (Fig. 3). Additional les show these in more detail (see Additional le 2-5). Additionally, the differentially expressed genes of HCC patients in the GSE57957 dataset were analyzed by the LIMMA packet. After comparison, a total of 373 differentially expressed genes were selected, including 88 upregulated genes and 285 downregulated genes (Fig. 4). Additional les show these in more detail (see Additional le 6-8). The cutoff criterion for gene screening was |LogFC|>1 with a P-value <0.05. Based on these data, a combined analysis was performed between methylation groups and transcription groups. The results showed that there were 12 genes with upregulated methylation and downregulated expression and 6 genes with downregulated methylation and upregulated expression existing in both datasets concurrently. A further WGCNA revealed that there were 4 overlapping genes: carbohydrate sulfotransferase 4 (CHST4), corticotropin releasing hormone binding protein (CRHBP), insulin like growth factor binding protein 3 (IGFBP3) and parathyroid hormone 1 receptor (PTH1R).

Genes affecting patient survival
We downloaded the mRNA expression data of HCC patients from the TCGA database. A total of 369 HCC patients with complete survival information and expression pro les were included. The survival status of patients with HCC under the in uence of each key gene is shown in Figure 6. A KM plot analysis revealed that CHST4 (p=1.534e -2 ), CRHBP (p=9.793e -3 ) and IGFBP3 (p=1.373e -2 ) signi cantly affected the survival of patients, while no signi cant association was found between PTH1R and patient survival (p=1.463e -1 ). Next, we focused on exploring the potential mechanism of action among these key genes in HCC patients and describing the molecular map of HCC-related multiomics, which will contribute to the treatment and management of HCC and improve the prognosis of patients.

Key genes in uencing tumor immune invasion
The tumor microenvironment is a functional whole that is mainly composed of tumor-related broblasts, immune cells, extracellular matrix, various growth factors, in ammatory factors, special physical and chemical characteristics and tumor cells themselves. The occurrence, growth and metastasis of tumor cells are closely related to their internal and external environment, which signi cantly affects the diagnosis, survival outcome and clinical treatment sensitivity of tumors. Therefore, we analyzed the relationship between the core genes obtained from the TCGA database that are involved in tumor immune invasion to further explore the potential molecular mechanism underlying the in uence of the core genes on the progression of HCC. The results revealed that three key genes were signi cantly correlated with the content of immune cells (Fig. 7). Among them, CHST4 was positively correlated with neutrophils, M0 macrophages and resting dendritic cells but negatively correlated with activated NK cells, CD8 T cells and gamma delta T cells; CRHBP was positively correlated with naïve B cells, plasma cells, and resting memory CD4 T cells, along with a signi cant negative correlation with M0 macrophages, memory B cells and memory CD4 T cells; and IGFBP3 was signi cantly positively correlated with follicular helper T cells, neutrophils and M0 macrophages but negatively correlated with monocytes, M1 macrophages and M2 macrophages.
. Key genes affect chemotherapeutic drug sensitivity Our study further explored the relationship between the core genes and commonly used chemotherapy drugs through chemotherapy sensitivity data from the GDSC database. The regression relationship between each chemotherapeutic drug and each core gene is shown in Figure 8. The results showed that high expression of the CHST4 gene could signi cantly increase the estimated IC 50 of erlotinib, AMG.706, CCT007093 and MG.132. In contrast, a general decrease in the estimated IC 50 of all six drugs was observed along with high expression of CRHBP. In addition, the high expression of IGFBP3 signi cantly improved the estimated IC 50 of the ve chemotherapeutic drugs except CCT007093. This nding suggested that core genes may dominate the prognosis of HCC patients by affecting sensitivity to chemotherapy drugs.

Mutations and signaling pathways involved in key genes
We selected the Pan Cancer dataset from the TCGA database and studied the mutations of core genes by the cBioPortal database. As a result, core gene mutations were present in the tumor cells of 32 HCC patients (9%). The speci c mutations of each core gene are shown in Figure 9. The mutation rates of the 3 key genes were 2.9% (CHST4), 2.0% (CRHBP) and 5.0% (IGFBP3). Only high mRNA expression was found in the CHST4 gene, while missense mutations, ampli cation and high mRNA expression occurred in the IGFBP3 gene simultaneously. Next, we studied the speci c signaling pathways involved in the three core genes to explore the potential molecular mechanisms by which key genes in uence tumor progression. GVSA showed that high expression of the CHST4 and IGFBP3 genes was associated in various signaling pathways, such as bile acid metabolism, fatty acid metabolism and DNA repair, while high expression of the CRHBP gene is associated with the PI3K-AKT-mTOR signaling, oxidative phosphorylation and glycolysis pathways, which are closely related to the occurrence and progression of HCC (Fig. 10). The GeneMANIA results suggested a close association among CHST4, ICRHBP, and IGFBP3 and related genes in various aspects involved in tumor progression, including the regulation of the insulin-like growth factor receptor signaling pathway, positive regulation of activated T cell proliferation, growth factor binding, smooth muscle cell (SMC) migration and its regulation.

Discussion
The development of HCC is a complex biological process characterized by the accumulation of genetic variation and epigenetic modi cation abnormalities. DNA methylation is an important part of the epigenetic regulatory mechanism in the human genome and responsible for regulating the expression and silencing of target genes by adding or removing a methyl group to the promoter regions under the catalysis of DNA methyltransferase [10]. Previous studies have con rmed that abnormal DNA methylation is common in human tumor cells, including HCC [2,11,13,14]. Oncogene demethylation and tumor suppressor gene methylation were found to be the main initiators and promoters in the progression of malignant tumors [15]. Currently, there is little information on the mapping of HCC methylation genes; therefore, our goal was to screen and validate the key differentially expressed methylation genes in HCC patients.
In this study, we extracted 78 groups of transcriptional data from the NCBI GEO public database. The WGCNA network and TOM analyses were performed, and the module with the highest correlation to tumor phenotypes was selected. GO and KEGG analyses suggested that the module was associated with several signaling pathways that have been veri ed to be closely related to the progression of HCC, including collagen-containing extracellular matrix, complement activation, lectin pathway, blood vessel development and cytokine receptor binding [16,17]. A total of 373 differentially expressed genes were screened from the module, including 88 upregulated genes and 285 downregulated genes. After a combined analysis, we nally identi ed four overlapping genes: CHST4, CRHBP, IGFBP3 and PTH1R. A KM plot analysis showed that three of the four genes (except PTH1R) signi cantly in uenced the predicted survival of HCC patients, which were collectively referred to as key genes. It is well known that multiple immune-mediated pathways are involved in the development of tumor microenvironments. The tumor microenvironment contains various components that can inhibit the growth and proliferation of tumor cells, including immune effector molecules and cells, and it also has components that promote the growth and metastasis of tumor cells, such as tumor-associated macrophages and immunosuppressive molecules. Therefore, the relationship between the three key genes and tumor immune invasion was also analyzed. The results indicated that each gene had a strong correlation with immune cells, including M0 macrophages, neutrophils, and naïve B cells. A drug sensitivity analysis suggested that gene overexpression could signi cantly affect the lethality and sensitivity of commonly used chemotherapy drugs to tumor cells, thus further producing a chain reaction on the prognosis of HCC patients. Moreover, we found that high expression of the CRHBP gene can enhance the targeting sensitivity of chemotherapy drugs, which is contrary to the effect of CHST4 and IGFBP3. In addition, a total of 32 HCC patients (9%) were con rmed to have at least one key gene mutation. GSVA suggested that the three core genes participated in multiple speci c signaling pathways involved in liver metabolism and tumor progression, such as bile acid metabolism, DNA repair, oxidative phosphorylation and glycolysis [18,19]. Their involvement in T cell proliferation, growth factor binding and SMC migration has also been revealed by GeneMANIA. These results demonstrated the roles of CHST4, CRHBP and IGFBP3 as key differentially expressed methylation genes in HCC.
CHST4 is responsible for encoding N-acetylglucosamine-6-O-sulfotransferase 2 (GlcNAc6ST-2), which is composed of 386 amino acids. GlcNac6ST-2 can participate in the synthesis of L-selectin ligand and regulate the capture and rolling of white blood cells on vascular endothelial cells, thereby mediating the initial interaction between lymphocytes and high endothelial venules (HEVs). HEV is a special type of postcapillary venule characterized by lymphocyte homing. In normal human tissues, CHST4 expression can be detected in the intestinal tract, pancreas, and liver [20]. When in ammation is present, HEV-like blood vessels and GLCNA6ST-2 expression will appear at the source, such as the lungs of asthmatic sheep and the synovium of patients with rheumatoid arthritis [21,22]. The expression and function of CHST4 in tumor cells have also been found and con rmed in previous studies. Among all tumors, cholangiocarcinoma has the highest frequency of CHST4 gene change [20]. In colorectal adenocarcinoma, the transcriptional expression of the CHST4 gene was signi cantly higher than that in normal tissue [20]. Accelerated leukocyte in ltration occurred in CHST4 knockout mice, thus con rming the barrier protective role of CHST4 in colonic mucosa [23]. The overexpression of GlcNAC6ST-2 was also found in early mucinous ovarian adenocarcinoma and clear cell carcinoma, and it was not observed in benign ovarian tumors. The mutation and high expression of CHST4 were also found in gastric cancer tissues and gliomas, suggesting a poor prognosis [24,25]. In HCC, low mRNA expression and high protein expression of the CHST4 gene have been detected compared with normal tissues in previous studies [20]. Notably, higher CHST4 expression in HCC patients indicated poor overall survival (OS) and disease-free survival (DFS), which is consistent with our ndings, whereas lower CHST4 expression in HBV-induced HCC patients indicated poor OS and DFS. These trends may be the result of low CHST4 expression promoting HBV expression and replication in liver cells in the absence of immune system defense[26].
Moreover, CHST4 overexpression has been proven to impede the migration ability of HCC cells [20].
CHST4 gene downregulation disrupts the immune response, leading to neutrophil in ltration and microvascular angiogenesis, which increases the risk of extrahepatic metastasis [20]. Similarly, the expression of CHST4 was signi cantly positively correlated with the degree of in ltration of immune cells, including B cells, CD4 cells, and macrophages, which was con rmed in our study but had no correlation with age, sex, TNM stage or serum AFP [20]. Racial differences in CHST4 gene expression have also been observed, which may explain the bias in gene expression from different samples. Although the association between CHST4 gene expression and HCC progression has been partially studied, there are few studies on CHST4 methylation in HCC.
Together with urocortin, urocortin 3 (UCN3), and corticotropin releasing hormone (CRH) receptors 1 and 2 (CRHR1 and CRHR2), CRHBP is classi ed as a member of the CRH system, and it is primarily responsible for regulating the expression balance between CRH and CRHR1/CRHR2 by binding to the CRH complex. The mechanism of CRHBP gene methylation in tumor cells has been explored in several previous studies, especially in urinary tumors. A signi cant reduction in mRNA levels caused by CRHBP gene methylation in renal cell carcinoma (RCC) has been demonstrated in several studies [27,28], and it has also been observed in bladder and prostate cancers [28,29]. Tezval et al. observed an approximately 80-fold increase in the CRHBP CpG island sequence leading to hypermethylation in over 50% of RCC tissue samples [28]. He et al. reported similar data in another article and showed that CRHBP mRNA levels in clear cell carcinoma were 33-fold reduced compared with those in normal tissues [27]. In addition, the use of methyltransferase inhibitors increased the mRNA reexpression level by approximately 100 times and reduced the aggressiveness of RCC, suggesting the reversibility of methylation and the possibility of tumor therapy [28]. At the same time, the degree of gene methylation was signi cantly positively correlated with distant metastasis and advanced disease. Lower mRNA expression of CRHBP predicted shorter survival and poorer prognosis. The mechanism may be that low expression of CRHBP leads to decreased activation and loss of function of CRHR2. CRHR2 mediates tumor cell apoptosis and inhibits vascular endothelial growth factor (VEGF) expression by competitively binding CRH with CRHR1, thus leading to reduced angiogenesis and a lower risk of tumor metastasis and dissemination [30]. To some extent, the decrease in CRHR2 permeability in tumor microvascular endothelial cells and the increase in CRHBP overexpression in RCC apoptosis also con rmed this mechanism. Furthermore, indicators of low CRHBP expression have been found in ovarian and breast cancer cells. Other members of the CRH system have also been reported to be involved in tumor progression. For example, CRH was detected to promote the growth and proliferation of colon cancer cells by activating CRHR1 and upregulating VEGF expression [31]. Tabrizi et al. also reported that RCC samples presented low levels of UCN3 mRNA expression, which inhibits angiogenesis by activating CRHR2 [30]. In the eld of HCC, there are few reports on the methylation and mechanism of the CRHBP gene. Xia et al. con rmed that the protein expression of the CRHBP gene was signi cantly reduced in HCC tissues compared with adjacent normal tissues [32]. Additionally, low expression of the CRHBP gene was signi cantly associated with the Edmondson grade, high serum AFP and tumor size. Notably, there is a close association between hypermethylation of the CRHBP gene and long-term stimulation of HBV [32], although the mechanism is unclear. Finally, CRHBP is considered a tumor suppressor gene because of its ability to activate p53-mediated tumor apoptosis and promote tumor immunity by the NF-kB signaling pathway [33]. A breakthrough in the methylation of the CRHBP gene will be a new focus for the treatment of cancer in the future.
The insulin-like growth family consists of insulin-like growth factors (IGF1 and IGF2), IGF receptors (IGFRs) and six IGFBPs (IGFBP1-6). IGFBP3 is the most widely and deeply studied IGFBP species in vivo. It is produced mainly by Kupffer cells in the liver but is also found in other local tissues, including the breast, lung, and digestive tract. It has been reported that IGFBP3 can reduce the pro-proliferation and anti-apoptotic effects of IGF/IGFR in cells and control the bioavailability of IGFs by binding to IGF with a high a nity [34]. Moreover, the inhibitory effect of IGFBP3 on angiogenesis was also demonstrated. When any link in the axis is damaged, the mechanisms involved in tumor development, chronic in ammation and diabetes are activated [35]. The methylation of IBFBP3 in human tumor cells has been widely reported over several years. An association between high expression of IGFBP3 and poor prognosis in breast cancer patients was found by some researchers[36]. Zielinska et al. observed that impaired glucoseregulated protein 78 (GRP78) synthesis in breast cancer resulted in the inability of IGFBP3 to enter cells, where IGFBP3 plays a pro-cell proliferative role by binding to cell surface factors [37]. This nding is consistent with the worst OS in GRP78-negative/IGFBP3-positive patients. The hypermethylated expression of IGFBP3 in gastric cancer and colorectal cancer has been proposed and veri ed by several reports [38]. In addition, Adachi et al. pointed out that IGFBP3 overexpression combined with adjuvant chemotherapy could inhibit tumor proliferation mediated by the NF-kB pathway, which would be bene cial for the prognosis of patients with gastric cancer [38]. The combined diagnosis of CA19-9, IGFBP2 and IGFBP3 has also been advocated to improve the detection rate of early pancreatic cancer [39]. High levels of IGFBP3 methylation in lung, ovarian, endometrial, and prostate cancers were mentioned to varying degrees [35,40]. In contrast, overexpression of IGFBP3 in nasopharyngeal carcinoma was reported to have precisely the opposite effect, thus promoting proliferation, invasion, and metastasis of cancer cells. Similar to most other tumors, IGFBP3 expression was observed to be signi cantly lower in HCC than in adjacent nontumor tissues [34]. This reduction is partially associated with insu cient secretion from damaged liver cells, degradation of proteases, and hypermethylation of genes. The CpG promoter region of IGFBP3 has been proven to have binding sites for the tumor suppressor gene p53 [34].
Methylation of this region inhibits p53 gene-mediated apoptosis. At the same time, reduced IGFBP3 secretion will encourage free IGF to bind to IGFR, which plays a role in promoting tumor cell proliferation. However, Bai et al. reported the opposite conclusion: high IGFBP3 expression was associated with poor prognosis [41]. In some studies, activation of IGFBP3 has been found to reduce cell proliferation and promote apoptosis [34], which may provide a direction for new targeted therapies in the future.
In conclusion, this study proposed a method of screening for key differentially expressed methylation genes in HCC patients. By discussing and verifying the preselected genes, we nally determined the core roles of CHST4, CRHBP and IGFBP3 in HCC gene methylation. Our work helps to clarify the role of gene methylation in the progression of HCC and shows that it is involved immune in ltration, molecular mechanisms and drug sensitivity. These results may lead to breakthroughs and possibilities for early accurate diagnosis and effective targeted therapy for HCC in the future.
Weighted Gene Co-expression Network Analysis and Differential Gene Expression Analysis. Front Genet 2020, 11:615308.   Heat map of methylation differences between the normal group and HCC group.